Chapter 3: Risk analysis and apportionment in construction The research literature on risk analysis and apportionment in construction, and formal and analytical techniques proposed in the literature for contractor’s risk analysis at the tender stage of projects are reviewed. This Chapter covers the analytical prescriptions that seek to propose how contractors should approach risk when bidding. 3.1 Risk in relation to contractors Little attention has been focussed on a precise definition and evaluation mechanism for project management risk specifically related to contractors. Project Management Institute (2004: 238) define risk as an uncertain event or condition that, if it occurs, has a positive or a negative effect on at least one project objective, such as time, cost, scope, or quality. However, three risk definitions taken from both the construction management and finance literature may help to better understand risk in the context of contractors. First, in developing a systematic influence diagram- based model for contractors risk analysis at the tender stage, Al- Bahar and Crandall (1990: 534) described risk as "an exposure to the chances or occurrences of events adversely or favourably affecting project objectives as a consequence of uncertainty". Second, a practitioners’ textbook prepared on behalf of the Aqua Group by Hackett et al. (2007: 35) defined risk as "the possible loss resulting from the difference between what was anticipated and what finally happened". Third, a financial analysis and management textbook by Fisher and Jordan (1996: 70) defined risk as "the possibility that realized returns will be less than the returns that were expected". Thus, risk may be understood in the context of contractors as a positive or negative deviation to expected profit. This aligns with Chang and Tien’s (2006: 171) definition of risk as "a measure of the probability and consequence of not achieving a goal." Risk is not the same as risks, although the terms are often used interchangeably in the literature. Whiles risk is the deviation to an expected outcome, risks are the actual deviation-causing events. As explained in a financial analysis textbook by Fisher and Jordan (1996: 70), forces [risks] that contribute to variations in return constitute elements of risk. Al-Bahar and Crandall (1990) defined risk event as what might happen in favour or in detriment of a project. In examining the way software practitioners are taught to perform risk management, Pfleeger (2000: 266) stated three criteria for identifying a risk event. First, a loss associated with the event, often called the risk impact. Second, the likelihood that the event will occur, with risk probability often measured with a number between zero (impossible) and one (certain). Third, the degree to which the project team can change the outcome, either by mitigating the risk’s causes before they occur or by controlling the risk's effects afterwards. An experiential- based textbook by Park (1979: 170) explained 12 risk events contractors face as: weather, unexpected job conditions, personnel problems, errors in cost estimating/scheduling, delays, financial difficulties, strikes, faulty materials, faulty workmanship, operational problems, inadequate plans or specifications, and disaster. 3.2 Types of risk in construction contracts Risks in construction have been classified in different ways (see for example, Edwards and Bowen’s (1998: 341-344) comprehensive review of risk literature (1960-1997) in construction). However, they significantly have the same meaning in that authors generally agree that some risks can be controlled whereas others cannot. Murdoch and Hughes (2008: 81) classified risks affecting construction projects under physical works, delay and disputes, direction and supervision, damage and injury to persons and property, external factors, payment, and law and arbitration. Erikson (1979: 6) classified risks in construction as contractual risk (caused by lack of clarity, absence of communication between parties, problems of timeliness in contract administration) and construction risk (inherent in the work itself). In developing a fuzzy model for contractor’s risk assessment at the tender stage, Tah et al. (1993: 282) categorized project risks into external and internal risks (see below). This is similar to the classification in the finance literature where portfolio theory and capital-market theory divides risk into systematic risk (external – overall market risk including unanticipated increases in inflation or interest rates, labour shortages, and economic downturn or recession) and unsystematic risk (internal – independent of any economic, political, or social factors which affect the market in a systematic way, including the risks mentioned by Park, 1979). 3.2.1 External risks External risks are those that are prevalent in the external environment of projects, such as those due to inflation, currency exchange rate fluctuations, technology change, major client induced changes, politics, and major accidents or disasters. They are relatively non-controllable and so there is the need to continually scan and forecast these risks and in the context of a company’s strategy (Tah et al., 1993). 3.2.2 Internal risks According to Tah et al. (1993), internal risks are relatively more controllable and vary between projects. They include the level of resources available, experience in the type of work, the location, and the conditions of contract. Some of these risks are local to individual work packages or categories within a project, whilst others are global to an individual project and cannot be associated with any particular work package. The local risks cover uncertainties due to labour (availability, quality, and productivity), plant (availability, suitability, and productivity), material (availability, suitability, supply, wastage) and subcontractor (availability, quality, productivity, and failure) resources and the site (ground conditions, accessibility, type of work, complexity of work). They are considered for each work package in the case of bill of quantities. Global risks are often allocated to the project as a whole because of their very nature. They cover risks relating to the performance (management experience, availability of partners, relationship with client, workload commitment), contract (contract type, contractual liabilities, amendments to standard form), location (head office, project) and financial (cashflow, funding, economic conditions) aspects of the project. 3.3 The PMBOK guide to project risk management The Project Management Institute (2004: 237-268) covers project risk management (PRM). The objectives of PRM are to increase the probability and impact of positive events, and to decrease the probability and impact of events adverse to the project. The risk identification process, which usually leads to the qualitative risk analysis process, is an iterative process of determining which risks might affect the project and documents and their characteristics. The PMBOK guide also outlines inputs, tools, and techniques that may be used to identify and quantify risks. According to the guide, PRM includes the processes concerned with conducting: Risk Management Planning (deciding how to approach, plan and execute the risk management activities of the project); Risk Identification (determining which risks might affect the project and documenting their characteristics); Risk Analysis (see below); Risk Response Planning (developing options and actions to enhance opportunities, and to reduce threats to project objectives); and Risk Monitoring and Control (tracking identified risks, monitoring residual risks, identifying new risks, executing risk response plans, and evaluating their effectiveness throughout the project life cycle). The primary outputs from a risk identification exercise may be entered into a risk register, which typically contains: a list of identified risks; list of potential responses; root causes of risk; and updated risk categories. Risk analysis involves two main aspects as follows. 3.3.1 Qualitative risk analysis Qualitative Risk Analysis covers the methods for prioritizing identified risks for subsequent further analysis or action by assessing and combining their probability of occurrence and impact. The tools and techniques for qualitative risk analysis include (1) risk probability and impact assessment; (2) probability and impact matrix; (3) risk data quality assessment; (4) risk categorization; and (5) risk urgency assessment. The inputs required for a qualitative risk analysis are (a) organizational process assets; (b) project scope statement; (c) risk management plan; and (d) risk register. 3.3.2 Quantitative risk analysis Quantitative Risk Analysis – numerically analysing the effect on overall project objectives of identified risks that have been prioritized by the qualitative risk assessment process. The inputs/information required for a quantitative risk analysis are (1) organizational process assets; (2) project scope statement; (3) risk management plan; (4) risk register; and (5) project management plan. The tools and techniques for quantitative risk analysis include (a) data gathering and representation techniques – interviewing, probability distribution, and expert judgement; (b) quantitative risk analysis and modelling techniques – sensitivity analysis, expected monetary value analysis, decision tree analysis, and modelling and simulation techniques. 3.4 Risk pricing in economics and finance Economists and financial analysts have traditionally relied on the capital asset pricing model (CAPM) of Sharpe (1964), Lintner (1965) and Mossin (1966) to price risk. One study that relates the application of portfolio and capital market theory to construction businesses is the one of Farid et al. (1989). Given certain simplifying assumptions, the CAPM states that the rate of return on any security is linearly related to that security systematic risk (or beta) measured relative to the market portfolio of all securities. The CAPM relies on many assumptions. For example, that investors have the same information, and this information is costless to gather and process. In addition, there should be no taxes, transaction costs, or other frictions. It also assumes that investors can borrow and lend at the risk-free rate of interest. The model takes into account of the tendency for an asset to covary with the market portfolio, and represents the non-diversifiable risk that investors are compensated for bearing, i.e. the unsystematic risk. The CAPM is appealing in its elegance and logic. However, empirical tests of the model often raise doubts about its underlying assumptions. For example, Brennan (1970) and Litzenberger and Ramaswamy (1979) relaxed the “no tax” assumption. Breeden (1979) and Merton (1973) made further extensions to the CAPM. Ross (1976) developed an entirely new model called the Arbitrage Pricing Theory. Clearly, the model’s focus on the market rather than total risk is a good way of thinking about the usefulness of assets in general. But it is unknown precisely how to measure any of the inputs required for using the CAPM. The inputs should all be ex- ante, yet usually only ex-post data is available. Although it may be useful in the finance industry, when we turn to construction, there is often little historical data on projects. This situation has been the foundation for the calls to focus risk assessment models on the linguistic data that can more readily be obtained from experts and people with the relevant knowledge (Kangari and Riggs, 1989; Al- Bahar and Crandall, 1990). Outputs of the CAPM are often subject to potentially large errors since historical data for beta coefficients (the actual measure of the risk) vary depending on the time period and the methods used to estimate them. The finance sector has advanced in the way that risk is managed over the past two decades. According to Dubofsky and Miller (2003), one of the methods for managing risk other than just the beta coefficient of portfolios is the use of derivatives. The authors explain that a derivative is a financial contract whose value is ‘derived from’ or depends on the price of some underlying asset. They go on to say that the value of a derivative changes when there is a change in the price of the underlying related assets. Four of these instruments are defined here: forwards, futures, swaps and options. A forward contract gives the owner the right and the obligation to buy a specified asset on a specified date at a specified price. The seller of the contract must then sell the asset on the date for that specified price. At delivery, ownership of the good is transferred and payment is made. Thus, whereas a forward contract is originated today, and the price is agreed upon today, the actual transaction in which the underlying asset is traded does not occur until a later date. Second, a futures contract has a similar arrangement except that rights under such an agreement can be transferred to third parties, and therefore futures are traded on organized exchanges and are more liquid than forwards. Third, swaps are contracts between two parties to exchange cash flows at future dates according to prearranged formulae. Fourth, an option is a contract entitling the holder to buy or sell a designated security at or within a certain period at a particular price (Rubenstein 1983, Brennan 1986). However, due to the sophistication of the finance sector, practitioners of the construction sector often use mechanisms such as a fixed single percentage or an overall lump sum based on intuition and experience to price for the risk in construction projects (Baloi and Price, 2003). Similar to shortcomings in the application of the CAPM, Williams (1996) exposes the inaccuracies, despite a general acceptance, of reducing a wide range of potentialities and contingencies down to one measure of project management risk as a function of probability multiplied by impact. Unsurprisingly, the Project Management Institute (2004: 237-268) also advocates the application of this approach (and a probability and impact matrix) for the quantitative assessment of project risk. The model may not predict accurate forecasts of risk. However, its use is still encouraged in practice. 3.5 Empirical studies on how contractors approach risks The literature on how contractors approach risk in their pricing of work is reviewed. 3.5.1 Risk and price relationship The following four cases in the construction management literature show that risk apportionment could influence prices by up to 17% of a bid price. First, a bid simulation experiment by Neufville and King (1991) in which 30 US contractors were involved showed that contractors add significant premiums, around 3% of the total value of a project, to their bid markups to compensate for risk and their lack of enthusiasm to do a job. However, there was no account of how contractors behave when they do indeed need the work. The second case is an interview study of 30 specialist contractors in the US by Shash (1993) showed that they would generally increase their prices by 5-10% if they were uncertain about a main contractor with whom they had no previous experience. The third case is represented by another interview study of 12 US contractors by Smith and Bohn (1999) showed that risk analysis generally has no impact (0%) on bids during times when contractors have a high need for work and competition is high. In the fourth, Atkinson (2007) reported the case1 involving problems caused by risk on a specific design and build project won by Balfour Beatty – the final link of the M60 outer ring road around Manchester from Medlock to Irk. The works involved cut-and-fill of 2 million m3 of excavation and 1.8 million m3 of deposition. Blackwell, who won the fixed-price job to carry out the earthworks under a subcontract to Balfour Beatty, performed a formal risk analysis and included 17% of the subcontract price for unseen ground conditions and weather. Therefore, risk analysis and apportionment may have a significant impact on tenders. Another area of literature is about factors that influence contingencies. Several surveys have been carried out on this topic. In an interview of 30 US contractors, Neufville and King (1991: 664) found that the following factors affect bidding contingencies: project complexity, identity of client, quality of design, identity of consultants, site conditions, logistics, project duration, and safety hazards. On the same topic, Smith and Bohn (1999: 106) interviewed 12 US contractors and identified the following factors: workload, contract size, project complexity, number of bidders, owner’s reputation, bidder’s mentality, clarity of contract documents and bidding period. A study of 38 contractors in Hong Kong by Wong and Hui (2006: 431) showed that project characteristics, client’s identity, consultant’s identity, contractor- related issues, contract documents, contract administration, bidding situation and economic environment affect contingencies. The findings are significantly similar, especially given the time elapsed between the different studies. 1 CA Blackwell (Contracts) Ltd v Gerling Allgemeine Versicherungs (2007) EWHC 94 LAWTEL Commercial Court Table 3.1 Empirical studies in journals on how contractors price risk Authors Yea Journ Vol Issu Pages Research Data Country r al . e method poin ts Uher 199 CME 9 6 495- Q. Survey 20 Australia 1 508 Neufville and 199 JCME 11 4 659- Exp./Intervi 30 US King 1 7 673 ew Mak and Raftery 199 CME 10 4 303- Q.&I Survey 25 UK 2 320 Tah et al. 199 CME 12 1 31-36 Q.&I Survey 7 UK 4 Akintoye & 199 IJPM 15 1 31-38 Q. Survey 30 UK MacLeod 7 Bajaj et al. 199 CME 15 4 363- Q Survey 19 Australia 7 369 Asaaf et al. 200 IJPM 19 5 295- Q. Survey 61 Saudi 1 303 Arabia Wong and Hui 200 CME 24 4 425- Q. Survey 38 Hong 6 438 Kong Chan and Au 200 IJPM 25 6 615- Q. Survey 60 Hong 7 626 Kong CME: Construction Management and Economics; IJPM: International Journal of Project Management; JCEM: Journal of Construction Engineering and Management 3.5.2 Risk pricing mechanisms of contractors Table 3.1 shows empirical studies related to how contractors price risk when bidding. The Table of literature shows that there are few studies related to how contractors actually price risk in bids, and these are based on questionnaire and interview surveys of contractors in five different countries. The literature review identified nine methods that contractors use to price risk in bids. These methods were identified in studies by Neufville and King (1991: 665) and a discussion of Liu and Ling’s (2005) fuzzy logic- based artificial neural network markup estimation model by Connolly (2006: 657). Two are described in the former study, three in the latter. Neufville and King’s experimental and interview-based study of 30 contractors in the US showed that the contractors quantified risk by comparing the amount of labour they carried out on a project to the total project cost and varied their markup based on the size of this ratio. The two main methods they used to compensate for risk in bids were: • To develop a standard cost estimate (not considering risk) before varying the markup for risk. • To develop a cost estimate (that adjusts productivity factors or contingencies based on the risk of each estimated item) before applying a standard markup to the risk-compensated estimate. Connolly (2006) used his practical experiences to describe three of the ways in which contingencies are priced in practice as: • The cost of the works (calculating the work quantities (materials, labour and equipment) before using a historical database to calculate a margin of error around the accuracy of the quantities. A probability range is then used to decide a contingency on the cost of works). • The cost of risk (the risk attributes can alternatively be analysed using probability functions or another approach). • The price of profit (allocating a contingency for the risk). A survey of seven UK contractors by Tah et al. (1994) showed that contractors may apportion bidding contingencies as: • A percentage in the profit margin. • A separate percentage in all the costs. • A lump sum in the entire preliminary bill. • A percentage in one bill if the risk is in that bill alone. 3.5.3 Contractor risk management practices Risk management is mostly defined as a logical process of risk identification, risk analysis and evaluation, and risk monitoring and control (PMI, 2004). Contractors have often been portrayed to be poor at managing risk by for example, authors such as Baloi and Price (2003: 262), Ahmed et al. (2002: 4) and Kangari and Riggs (1989: 126). In developing an analytical model for modelling global risks, Baloi and Price (2003) said: "…many contractors are unfamiliar with these risk factors and do not have the experience and knowledge to manage them effectively. As a consequence, conflicts, poor quality, late completion, poor cost performance and business...Contractors have traditionally used high mark-ups to cover risk but as their margins have become smaller this approach is no longer effective...Contractors rarely use these techniques and tools in practice. More often than not construction contractors and other practitioners rely on assumptions, rules of thumb, experience and intuitive judgement which can not be fully described by prescriptive or normative models. Individual knowledge and experience, however, need to be accumulated and structured to facilitate the analysis and retrieval by others." According to Ahmed et al. (2002), "The construction industry has a poor reputation in coping with risks, many projects failing to meet deadlines and cost targets." Kangari (1989) said: "…the construction industry has a very poor reputation for coping with risk. Risk analysis is either ignored or done subjectively by simply adding a contingency. As a result many major projects fail to meet schedule deadlines and cost targets with attendant loss to both contractors and owners." However, these assertions may not be true generally. As showed in chapter one, contractors have had their own ways of dealing with risk since the early part of the 19th century (Hughes and Hillebrandt, 2003). A questionnaire survey of 19 contractors in Australia by Bajaj et al. (1997) identified five of the ways used by contractors for identifying risk at the tender stage of projects: • Risk review (by senior staff at the start of the tender pricing); • Contact (discussions with subcontractors, architect and client); • Research (ascertaining information about subcontractors, client, consultants, economic climate, etc); • Site visit (visiting site to ascertain the access situation, location, obstructions, etc); and • Finance (issues regarding payment and financial obligations). However, the study also exposes some of the weaknesses of questionnaire-based research: contractors were asked nine ‘yes’ ‘no’ questions about risk identification techniques that the researchers had in mind, apparently, before the study based on their knowledge and opinion. Eventually, the authors concluded that 53% of contractors use scenario building to identify risk, 5% (influence diagram) and 58% (questionnaire and checklist). However, none of these methods were actually indicated by any of the contractors in the five methods that they themselves had used to describe their risk identification procedures. Therefore, it is difficult to understand the extent to which the latter findings are valid given the study design. For example, it seems that little effort was made in isolating the measurement of the variables of interest from other effects. Hence, for example, it is not possible to determine the number of firms that are combining two or more risk identification approaches. Most of the firms routinely identify risks through the opinion of one or two experienced persons within the company. Under the above-mentioned risk identification mechanisms, contractors may also identify risks through the following five ways: circulating the tender information to persons in the company; using only the judgement of the estimator; a departmental tender review meeting; brainstorming; and using the opinion of external consultants. Table 3.2 Empirical studies in journals on risk management practices of contractors Authors Yea Journa Vol. Issu Pages Researc Data Country r l e h poin method ts Mok et al. 199 CME 15 2 161- Q. 52 Hong Kong 7 175 Survey Bajaj et al. 199 CME 15 4 363- Q. 19 Australia 7 369 Survey Akintoye and 199 IJPM 15 1 31-38 Q. 30 UK MacLeod 7 Survey Shen 199 IJPM 15 2 101- Q. 25 Hong Kong 7 105 Survey Tummala et al. 199 IJPM 15 5 297- Q. 20 Hong Kong 7 312 Survey Smith and Bohn 199 JCME 125 2 101- I. Survey 12 US 9 108 Kartam and Kartam 200 IJPM 19 6 325- Q. 35 Kuwait 1 335 Survey Ahmed et al. 200 JCE 17 1 4-11 Q. 30 US 2 Survey Lyons and 200 IJPM 22 1 51-61 Q. 44 Australia Skitmore 4 Survey CME: Construction Management and Economics; IJPM: International Journal of Project Management; JCEM: Journal of Construction Engineering and Management; JCE: Journal of Construction Engineering Table 3.2 shows published studies in journals on the ‘risk management’ practices of contractors. However, none of the authors actually proved that the contractors involved respond to risks according to the three formal, systematic processes outlined in the PMBOK (2004: 237-268). Given the emergence of the area, most authors seemed to have a preconceived idea of the theory that they simply confirmed through ‘ticks’ provided by respondents in the authors’ developed questionnaires rather than testing to see whether 'formal' risk management actually happens in practice. Akintoye and MacLeod (1997) used a questionnaire survey to elicit the opinions of 30 contractors and 13 project management consultants about risk perceptions and risk management practices in the United Kingdom. An open-ended question about the meaning of risk produced a wide variety of responses, but most respondents included some reference to the occurrence of adverse events, relating to time, cost or quality, in their descriptions of risk. Respondents were also asked to rate (from ‘high’ to ‘low’) the extent of risk premium they would attach to different risks. For both subgroups, financial risks rated highest, with legal and economic risks were close behind. Adverse weather, a natural risk, was rated lowly by both groups. As far as risk management techniques are concerned, about a third of contractor respondents used risk premiums (contingency allowances); about half used sensitivity analysis; while 20% of them knew about Monte Carlo simulation, only 3% ever used it. Project managers were more familiar with probabilistic techniques than contractors, and tended to use them more, but not by an enormously large degree. Tah et al. (1994), in a survey of the estimating practices of seven contractors, found that all of them classified risk as either ‘quantifiable’ or ‘unquantifiable’ (it is not clear how the survey instrument posed this question). For quantifiable risks, estimators included the costings in the estimate; while for unquantifiable risks an amount, based upon management perception of the situation, is added to the estimate either by increasing the profit mark-up or by including a lump sum in the preliminaries cost estimates. None of the respondents reported using statistical techniques to analyse risk. Using questionnaire, Mok et al. (1997) surveyed the risk perceptions and practices of 52 building services engineers in Hong Kong. They found that fewer than 30% of respondents possessed a comprehensive understanding of the meaning of risk; and fewer than 20% used probabilistic techniques of risk analysis. More than 75% of respondents most frequently used deterministic single-figure estimates, and more than 85% simply added a contingency sum to allow for risks. Less than 10% of respondents frequently used sensitivity analysis to check the validity of their risk allowances. On the other hand, more than 60% of respondents agreed that the establishment of probabilities constituted an inherent problem in quantitative risk analysis; while more than 50% agreed that interpreting the outcomes of risk analysis was difficult. 3.6 Formal and analytical risk models in construction Formal and analytical models developed to help contractors assess and price project risk at the tender stage have existed for 37 years now. Indeed, the first major journal paper identified in relation to risk pricing approaches for contractors was by Gates (1971). Table 3.3 shows a comprehensive literature of 67 analytical approaches that have been proposed in journals for contractor’s risk analysis at the tender stage. The Table also includes 11 seminal studies on risk analysis and management in construction which do not propose mathematical models per se. However, those studies appear to have informed the development of the mathematical contributions as most of them are cited by authors of the analytical models. The seminal studies include: a survey of 30 contractors and 13 project management practices on risk analysis and management in construction (Akintoye and MacLeod, 1997); a comprehensive review paper by Edwards and Bowen (1998) on future directions for risk management in construction; and an experimental/interview study by Neufville and King (1991) on the compensation for risk in bid markups by 30 US contractors. The literature here will form the basis of the theoretical data against which results of the empirical studies on how contractors take account of risks during tender preparation will be compared in order to identify whether there are areas of significant difference. The outcome will help in making recommendations for aligning theory and practice of how contractor price risks at the tender stage. The research method for achieving the first objective comprised of an examination of construction management journals, each from their first issue to articles in press (as at May 2008). The purpose was first to take account of all papers on risk that were published in them; and second, to log papers containing proposals for contractor’s risk analysis at the tender stage. A paper by Chau (1997) on “The ranking of construction management journals” provided a basic idea of journals in the field. The remaining papers were identified through a rigorous internet search which was followed by another search through a snowballing approach, namely references in papers identified earlier. The risk models identified and reviewed can be found in Table 3.3 Altogether, 67 analytical approaches for contractor’s risk analysis were found, and the frequency of publication proved to be increasing: 5 in 1970s; 11 in 1980s; 24 in 1990s; and 27 in 2000s so far. 3.6.1 Evolution of analytical risk models in construction Formal risk management, which is the systematic process comprising of risk identification, risk assessment (analysis and evaluation) and risk response (monitoring and control) evolved in the 1960s as a best practice theory for successful response to the risks in a project (Edwards and Bowen, 1998). However, a review of leading construction journals such as Construction Management and Economics (CME), International Journal of Project Management (IJPM) and ASCE Journal of Construction Engineering and Management (JCEM) shows that the proliferation of mathematical risk models actually started in the 1980s. The earliest analytical propositions were published in the American Society of Civil Engineers’ JCEM, for example Gates (1971), Vergara and Boyer (1977) and Ibbs and Crandall (1982). This journal was first of all known as the Journal of the Construction Division (JCD); the name was changed to Journal of the Construction Engineering Division (JCED), then to its current name. More than 25 of the models examined and classified were identified in this journal alone. The leading UK journals such as CME and IJPM both started being published in 1983. The first risk paper to be published in CME was the one by Beeston (1986). Although this paper did not introduce a mathematical model per se, its focus on showing how to combine risks in estimating probably generated interest in the topic and prompted the development of techniques that enhance the ability to deal with risks logically, for example Al-Bahar and Crandall (1990). Within the CME journal, more contributions on the topic of risk assessment and pricing came from authors such as Kangari and Riggs (1988) and Skitmore et al. (1989) but it was Birnie and Yates (1991) who first explored the potential of using techniques such as decision trees, utility theory and the Monte Carlo simulation technique to model risks in order to improve human judgement in construction cost prediction. In total, the journal has published over 20 risk models intended solely to help contractors estimate various forms of bidding contingencies. The first risk paper in the IJPM was the one by Barnes (1983). It proposed a risk-allocation algorithm. Cooper et al. (1985) and Baker (1986) made further contributions on risk response applications in construction. The former study assessed the reliability of the cost estimate for a large hydroelectric project and the adequacy of the contingency allowance. The latter presented a cost/time risk model for oil field development. Over 20 analytical risk models that are aimed solely to help contractors in estimating bidding contingencies were counted in the IJPM alone. Before the aforementioned UK journals were introduced in the 1980s, risk publications had long appeared in the much older periodicals of the American Society of Civil Engineers like JCEM for example. However, as Kangari and Riggs (1988) would note, the major application of portfolio theory to risk management in construction projects was by Vergara and Boyer (1977). However, prior to this, Gates (1967) had presented a model for quantifying, in monetary terms, contingencies for bidding mistakes, subjective and objective uncertainties, and chance variations. In his study, Dressler (1974) discussed how risk factors could be incorporated into the production sequence of segments of a linear site through a projected construction time, while Vergara (1974) supplied a model for monitoring changes in expected cost as a result of varying precision and variation in risk of a project as it progresses. There were further contributions by Spooner (1974) and Neufville et al. (1977). Altogether, over 25 risk models intended to help contractors in estimating various forms of bidding contingencies have been counted in the JCEM alone. These journals came long before others like Engineering, Construction and Architectural Management (ECAM), which was introduced in 1994. One seminal study in ECAM is a review paper by Edwards and Bowen (1998) on the future directions of risk management theory and applications in construction. Journals in which some other contributions were found include the Journal of Construction Procurement (JCP), CIB Journal of Construction Engineering (JCE) and Journal of Asian Architecture and Building Engineering (JAABE) that published a fault tree analysis-based risk analysis model for building projects by Tsai et al. (2002). 3.6.2 The fundamental approach for evaluating risk and its limitations Project management risk is fundamentally evaluated on the basis that one measure of a risk equals its probability times its impact (Turner, 1992). According to Zhi (1995), "…the risk concept is broken down into two main criteria: probability, and impact." However, there are significant limitations associated with the use of the fundamental PI risk model. According to Williams (1996), "Calculating ‘expected’ risk as probability multiplied by impact has limitations and ranking risks according to this figure is misleading. Computerized ‘risk lists’ thus ranked should not be relied upon. Both probability and impact must be considered at all times." Thus, it can be seen that the fundamental approach for taking objective account of project management risk is flawed in a way that Williams (1996) describes clearly. However, this theory is the basis of most analytical risk models. To date, the problem is still not resolved in the project management literature. To little avail, authors like Ward (1999), Williams (1993), Wynne (1992) and Charette (1989) tried to present alternative solutions which minimize the error associated with combining and comparing two or more risks according to the fundamental PI mode. Table 3.3 Analytical models for contractor’s risk analysis, and some seminal studies Authors Yea Journ Vol Issu Pages Modelling Category r al . e technique Gates 197 JCED2 97 2 277- Probability theory Classical 1 303 Spooner 197 JCD3 10 1 65-77 Monte Carlo / Prob. Classical 4 0 Dressler 197 JCD 10 4 571- Stochastic linear Classical 4 0 587 prog. Vergara 197 JCD 10 4 543- Probability theory Classical 4 0 552 Neufville et al. 197 JCD 10 1 57-70 Utility theory Classical 7 3 Carr 197 JCD 10 1 153- Utility curves Classical 7 3 161 Handa and 198 JCD 10 3 355- Probability theory Classical Georgiades 0 6 370 Warszawski 198 JCD 10 1 147- Probability theory Classical 2 8 157 Ibbs and Crandall 198 JCD 10 2 187- Utility theory Classical 2 8 200 Diekmann et al. 198 JCD 10 3 379- Quadratic Classical 2 8 389 programming Barnes 198 IJPM4 1 1 24-28 Algorithm Classical 3 Diekmann 198 JCD 10 3 297- Probability theory Classical 3 9 308 Ayyub and Halder 198 JCEM5 11 2 189- Fuzzy sets Conceptu 4 0 204 al Farid and Boyer 198 JCEM 11 4 374- Probability theory Classical 5 1 390 Beeston 198 CME6 4 1 71-79 Probability theory Classical 6 Baker 198 IJPM 4 4 205- Probability theory Classical 6 210 Franke 198 IJPM 5 1 29-34 Probability theory Classical 7 Carr 198 JCEM 11 1 151- Opportunity in Seminal 7 3 167 bidding Echeverry et al. 198 JCEM 11 1 1-18 Simulation Classical 8 4 Farid et al. 198 JCEM 11 1 109- RADR Classical 9 5 125 Kangari and Riggs 198 TEM7 36 2 126- Fuzzy sets Conceptu 9 131 al Williams 199 IJPM 8 2 84-88 Simulation Classical 0 Al-Bahar and 199 JCEM 11 3 533- Influence Classical Crandall 0 6 546 diagramming Seydel and Olson 199 JCEM 11 4 603- AHP Classical 0 6 623 2 ASCE Journal of the Construction Engineering Division 3 ASCE Journal of the Construction Division 4 International Journal of Project Management 5 ASCE Journal of Construction Engineering and Management 6 Construction Management and Economics 7 IEEE Transactions on Engineering Management Birnie and Yates 199 CME 9 2 171- Decision/risk Classical 1 186 analysis Neufville and King 199 JCEM 11 4 659- Risk and price Seminal 1 7 673 relation Diekmann 199 IJPM 10 2 75-80 Paradigm shift to Seminal 2 artificial intelligence Huseby and 199 IJPM 10 3 160- Influence Classical Skogen 2 164 diagram / Monte Carlo Russell and 199 CME 10 4 277- Four Classical Ranasinghe 2 301 moments/Pearson distribution Newton and Smith 199 CME 10 5 431- Non-deterministic Classical 2 449 Touran and Wiser 199 JCEM 11 2 258- Monte Carlo Classical 2 8 272 Tah et al. 199 CSE8 4 23 281- Fuzzy sets Conceptu 3 293 al Benny 199 IJPM 11 4 201- Simulation / Classical 3 208 probability Dey et al. 199 IJPM 12 1 22-33 AHP / Probability Classical 3 Moselhi et al. 199 JCEM 11 3 466- Neural networks Conceptu 3 9 479 al Paek et al. 199 JCEM 11 4 743- Fuzzy sets Conceptu 3 9 756 al Williams 199 IJPM 12 1 17-22 Probability theory Classical 4 Ranasinghe 199 CME 12 1 15-29 Uncertainty Classical 4 quantification Ranasinghe 199 CME 12 3 233- Two moments Classical 4 243 Gong and Rowings 199 IJPM 13 3 187- Simulation Classical 5 194 Zhi 199 IJPM 13 4 231- Probability theory Classical 5 237 8 Journal of Computing Systems in Engineering Table 3.3 Cont’d Authors Year Journa Vol. Issu Pages Modelling Category l e technique Mack 199 CME 13 5 385- Paradigm shift Seminal 5 392 from hard to soft approach Chau 199 CME 13 5 369- MC Simulation Classical 5 383 Williams 199 IJPM 14 3 185- Probability Classical 6 186 theory Akintoye and 199 IJPM 15 1 31-38 Survey Seminal MacLeod 7 Gong 199 IJPM 15 3 187- Probability Classical 7 192 theory Chapman 199 IJPM 15 5 273- Probability Classical 7 281 theory Dawood 199 CME 16 1 41-48 Probability Classical 8 theory Dawson 199 IJPM 16 5 299- Probability Classical 8 310 theory Edwards and Bowen 199 ECAM9 5 4 339- Review paper Seminal 8 349 Mulholland and 199 JCEM 125 1 8-15 HyperCard & Classical Christian 9 Excel Smith and Bohn 199 JCEM 125 2 101- Interview study Seminal 9 108 Chapman et al. 200 IJPM 18 1 337- Probability Classical 0 347 theory Wang and Huang 200 IJPM 18 2 131- Simulation Classical 0 138 Tah and Carr 200 CME 18 4 491- Fuzzy sets Conceptu 0 500 al Chapman and Ward 200 IJPM 18 6 369- Probability Classical 0 383 theory / Iteration Pender 200 IJPM 19 2 79-87 Paradigm shift Seminal 1 to real options Jaafari 200 IJPM 19 2 89-101 Paradigm shift Seminal 1 to strategy- based PM Kuchta 200 IJPM 19 5 305- Fuzzy sets Conceptu 1 310 al Hwee and Tiong 200 IJPM 20 1 351- Probability Classical 2 363 theory Hillson 200 IJPM 20 3 235- Paradigm shift Seminal 2 240 from risk to opportunity Patterson and 200 IJPM 20 1 365- Probability Classical Neailey 2 374 theory Ward and Chapman 200 IJPM 21 2 97-105 Paradigm shift Seminal 3 from risk to uncertainty Baloi and Price 200 IJPM 21 4 261- Fuzzy sets Conceptu 3 269 al Jannadi and 200 JCEM 129 5 492- Probability Classical Almishari 3 500 theory Nasir et al. 200 JCEM 129 5 518- Probability/BBN Classical 3 527 Zhong and Zhang 200 JCEM 129 5 501- Probability Classical 3 506 theory Choi et al. 200 JCEM 130 2 258- Fuzzy sets Conceptu 4 272 al 9 Engineering, Construction and Architectural Management Han et al. 200 JCEM 130 3 346- Probability Classical 4 356 theory Warszawski and 200 JCEM 130 3 357- Multifactor Classical Sacks 4 367 method Fang et al. 200 JCEM 130 6 862- Logistic Classical 4 868 regression Cioffi 200 IJPM 22 3 215- Differential Classical 5 222 equations Zeng and Ng 200 JCEM 131 2 176- Fuzzy sets Conceptu 5 186 al Lee 200 JCEM 131 3 310- Probability Classical 5 318 theory /Monte Carlo Liu and Ling 200 JCEM 131 4 391- Fuzzy sets / Conceptu 5 399 neural networks al Thomas et al. 200 CME 24 4 407- Fuzzy-fault Classical 6 424 tree/Delphi Poh and Tah 200 CME 24 8 861- Probability Classical 6 868 theory Ok and Sinha 200 CME 24 10 1029- Neural networks Conceptu 6 1044 al Diekmen 200 IJPM 25 5 494- Infl. diag./fuzzy Both 7 505 sets Zeng et al. 200 IJPM 25 6 589- Fuzzy sets / AHP Both 7 600 Ward (1999) examined the shortcomings of the PI technique, and said that for cost-effective management it is desirable to distinguish not only between the size of impacts and probability of impacts occurring, but also other factors such as the nature of feasible responses, and the time available for responses. He offered some practical suggestions for dealing with this problem. However, that was only the extent of the intervention – suggestions that others can attempt. Hence, the problem still remains. Williams (1993) offered the idea of plotting such risks on Probability-Impact Grids (PIG) in order to understand their effects. However, Wynne (1992) said four concepts should be distinguished: risk (where the odds are known), uncertainty (where the odds are not known but the main parameters may be), ignorance (where what we don’t know is not known) and indeterminacy (described as ‘causal chains or networks open’-so presumably implying an element of unknowability). Based on the different concepts proposed for approaching risk, Wright and Ayton (1994: 296) summarized and explained the approaches to risk perception and measurement as: experimental approach to risk; psychometric approach to risk; qualitative studies of risk perception; and alternative perspectives. 3.6.3 Risk analysis models proposed for contractors This section reviews some of the seminal ones among the 67 studies that typify the analytical propositions. This will form the basis of comparing theory and practice. Kangari and Riggs (1989: 127) categorized the risk analysis techniques as classical or conceptual. Classical risk modelling techniques, such as probability theory and Monte Carlo simulation, were used in developing most of the early project risk analysis models. However, in recent years, the paradigm shifted from "classicalism" towards "conceptualism" where conceptual modelling techniques like fuzzy sets and neural networks are increasingly being used for modelling risks in construction. The main reason for the paradigm shift was to help perform risk analysis using the more readily available linguistic subjective data, compared to objective historical data, which can be obtained from experts and persons with the relevant knowledge (Al-Bahar and Crandall, 1990). However, to date, there is no evidence showing that the paradigm shift has encouraged the application of risk models in practice. Neither has it appeared to influence much in theory since 35 of the 45 risk models from 1990 are still classical approaches. The generally accepted model for calculating one measure of project management risk is its probability multiplied by its potential impact (Al-Bahar and Crandall, 1990). This formula is widely used by academics and practitioners (Zhi, 1995). However, although still generally accepted, the practice of comparing or combining two or more risk events according to the fundamental formula of probability multiplied by impact has been criticized by authors such as Williams (1996) and Ward (1999). Two or more risk events that are evaluated using the model may indicate the same risk value but could have very different effects on a project on occurrence. The next section presents a detailed account on the fundamental model for evaluating project management risk, some insights into formal risk perception and measurement, and some seminal studies out of the 67 in Table 3.3. Beeston (1986) proposed a probabilistic method to help estimators in calculating risk premiums. The author described his method as: "…a practical way of building up an estimate which will be exceeded with a calculated probability." The method is based on the assumption that "an estimate for a project consists of a basic estimate for the project plus a list of itemized risks which may affect the cost…It is usual to state a sum of money for each item, referred to as risk allowances." To begin with, the author divided risk which should be priced in an estimate into two categories: fixed risk allowance; and variable risk allowance and described them as follows: "A fixed risk allowance is a sum of money which will either be incurred as a whole, with an estimated probability, or not at all [whiles] a variable risk allowance can occur to varying degrees so no fixed sum of money can be allocated to it." To cover the consequences of a fixed risk, the author prescribes the apportionment of an average risk allowance which is based on multiplying the fixed allowance with a probability of occurrence. Thus, the risk allowances for variable risk can be approached in two ways: (1) by breaking it down into several fixed risk allowances each with its probability of occurrence and apportioning a fixed risk allowance for each risk; or (2) by apportioning an average risk allowance based on an estimate of the sum of money which has which has a probability of 0.5 of being exceeded. However, the author acknowledged "…the difficulty which arises when it is necessary to consolidate the risk allowances to produce a combined risk allowance which can be added to the basic estimate…" Although all the maximum risk allowances can be totalled, the author describes this common approach as ‘too pessimistic’ since the chance of all risks occurring at this level is usually very low. Therefore, the author recommends a total risk allowance based on an assumption of ‘average luck’, i.e. totalling the average risk allowances and adding this ‘average risk estimate’ to the basic estimate for the project. A way of taking account of dependence is outlined as well as some views on risk allowance distribution. It can be seen that Beeston’s work represents a facile way of approaching risks in estimates. Given the lack of clear guidance on how probabilities should be assigned and the ‘average luck’ approach, it seems not to be straightforward. In developing the approach, there is no specific reference to any empirical work on what contractors actually do in practice. Citing a lack of significant work in construction risk analysis by fuzzy sets, Kangari and Riggs (1989) proposed a fuzzy set-based risk assessment methodology that can give contractors "…a more rational basis on which to make decisions." Following an analytical evaluations of risks based of fuzzy set principles, a risk value is generated which is recommended for inclusion the estimate as a risk premium. Although the authors thought the model could be useful in practice, they acknowledged the “sophistication and difficulty of applying the model in real life”. In short, this recognition probably meant that the technique is more appropriate for academia than practitioners. There was no reference to any empirical work on what contractors actually do underlying the model development, and no experiential knowledge is cited. Hence, it would appear that the proposal applies more to academia. Al-Bahar and Crandall (1990) introduced a systematic Construction Risk Management System for construction projects. In describing the purpose of the model, Al-Bahar and Crandall said: "The proposed model provides a formal, logical, and systematic tool that helps contractors in identifying, analyzing, and managing risks in a construction project. It is a logical substitute for the traditional intuitive unsystematic approach currently used by most contractors when they are dealing with risk." To quantify risks, the authors proposed three systematic steps for the risk analysis and evaluation process: data collection (comprising both objective statistical data, and subjective judgemental data); modelling uncertainty (comprising assessment of potential consequences, and assessment of probability distribution); and evaluation of potential impact of risk. This was described as the phase "…which incorporates uncertainty in a quantitative manner, using probability theory, to evaluate the potential impact of risk." Having earlier acknowledged what contractors tended to do in practice, an examination of the assumptions underlying the development of the CRMS model may indicate that the authors appear to be offering their model from the position that the intuitive and experiential-based approach used by contractors does not form an adequate professional and objective basis for serious project management decisions. Therefore, a more sophisticated and time-consuming procedure is offered. Here too, there is no specific reference to any empirical work on what contractors actually do in practice and there was no evidence cited to support the authors’ critique and sweeping generalization that contractors are unable to "systematically quantify risks and assess their potential consequences." Paek et al. (1993) developed a fuzzy set model to help contractors quantify project level risks when "…faced with the problem of deciding the bidding price of a construction project when the likelihood of the occurrence of risk events and the risk associated consequences are uncertain." The proposed risk-pricing algorithm consisted of two main stages: identifying the risk elements and quantifying the risk-associated consequences. Peak et al.’s model was based on the assumption that contractors traditionally determine a bidding price by calculating actual costs (direct, indirect and overhead costs) and adding a margin for contractor’s profit. Apparently, there is no reference to market premium which may, in fact, render the point of the whole risk analysis pointless after the Directors’ bid adjudication exercise. According to the technique, the risk elements of the contract should be first identified. Then, risk-associated consequences estimated as fuzzy numbers to reflect their uncertainty based on judgement of experts with relevant knowledge and experience and historical records that contractors experienced in past projects. Several experts estimate the cost or duration for each element. The most likely and least likely intervals of time or cost are determined from this data. The output represents the potential outcomes at a particular uncertainty level. Once all risk elements are summed, a ranking process can be used to determine an average value of the total potential loss. This loss represents the optimum contractor contingency which should be included in the bid. Given the sophistication involved, the authors admitted that such a complex mathematical model is warranted for only high-risk projects. Thus, the model may be used when bidding for very large projects where clients ask for a thorough professional risk analysis. A firm’s estimators can also use the model to assess the major risk elements and the monetary consequences associated with each risk element. The risk-pricing method was applied to an $800 million urban highway project to show how the computerized risk-pricing algorithm may be used. Moselhi (1995) discussed Paek et al.’s model by performing a program evaluation and review technique- (PERT-) like analysis, assuming a beta distribution to express the uncertainty associated with each element, on the same illustration used by Paek et al. As the discusser’s analysis yielded similar results, Moselhi asked two questions about the proposed approach: (1) what are the benefits of the proposed fuzzy method over the simple PERT-like approach; and (2) which of the two values presented by the models should a contractor have more confidence in and why? Tah et al. (1993) proposed a fuzzy set-based model that contractors can apply in subjective assessment of risks during tender preparation for the purpose of allocating contingencies to cover the risks. The authors justified the fuzzy set approach to the assessment of risks during tender preparation as follows: "In current practice, estimators add a large contingency to the overall estimate to compensate for inaccuracies and risks. This allocation is based mainly on the estimator’s perception of project risks. It is largely judgemental and arbitrary…company policy dictates a class of contingency which should be applied to different types of projects." However, there was no evidence shown to prove this assertion. The authors’ list of 16 references contained no comprehensive empirical study on the work of contractors. Thus, the assertion was more of a presumption, apparently. After presenting this view on what happened in current practice, the authors argued that "…this conventional method of contingency allocation [which contractors use in practice] may be argued to be too simplistic as it can easily reduce to a routine administrative procedure that requires little investigation and decision making by estimators and senior estimators." This was argued as the main reason for proposing the new methodology. The authors’ risk assessment model for contractor risk contingencies allocation at the tender stage was based on a hierarchical risk-breakdown structure (HRBS) which outlined three risk centres which should be considered in allocating risk contingencies. The contingencies were designated as C1, C2, and C3. Thus, the contingency for local risk, C1, should be derived as follows: m C1 = ∑ C1k (Eqn. 3.1) k =1 C1k are the individual local risks of the ‘m’ work packages within a given project. The model in Eqn. 3.1 proposes that allocation of C1 should be made at the level of the estimator. However, the allocation of the global risk C2 has to be a management decision, which depends on the management’s attitude to risk. The allocation of C2 is made at the level of senior management. The consideration of C3 for external risks was to be based on extraordinary occurrences. However, this was not supposed to be incorporated into the project cost estimates like the local and global risks. The reason is that corporate management would often have some reserve set aside for them or where contractors tender for major projects, there would be safeguard clauses to protect them from such extraordinary risks and transferred the risks to the client. Following an illustration of the model’s application to a hypothetical scenario, Tah et al. (1993) concluded that "The risk value helps to decide the percentage to apply to the risk-free direct cost estimate to cover local risks within the work package." Clearly, a close examination of this work would show that the analytical steps prescribed may be too cumbersome in reality. The work also brings forward three issues to be compared to results of the empirical studies: (1) whether the process used by contractors for pricing risk is deductive; (2) whether the direct cost estimate is indeed risk-free; and (3) whether contractors aggregate risks according to the rational sequence proposed in the model. In developing the approach, there is no specific reference to any empirical work on what contractors actually do in practice. Liu and Ling (2005) introduced a fuzzy logic-based artificial neural network model, to help contractors in the "...estimation of markup in a changeable and uncertain construction environment." The authors defined markup as "the sum of profit and contingencies." Therefore, the model was intended to help contractors in estimating a margin for both profit and risk. To begin with, the authors sought to "…investigate whether contractor’s markup estimation could indeed be accurately modelled” based on the premise that "...markup estimation has often been perceived as a kind of mysterious work, mainly based on the estimator’s intuition and experience, with some specific rules and constraints applied." No clear answer was provided about the question but the fact that the authors proceeded with the model development probably concluded that the markup function can be modelled. In justifying the model, the authors argued that "…it is important to be able to model markup estimation as the model can act as a decision aid to help contractors to overcome their shortcomings in judgement and limited short-term memory, which prevents them from processing large amounts of information". Here, the authors may appear to assume that the intuitive and experiential-based approach used by contractors does not form an adequate professional and objective basis for serious project management decisions. A postal questionnaire survey of 29 US contractors was carried out to "determine the most important and significant variables that affect markup estimation". However, the questionnaire itself seemed restrictive since 52 attributes, in seven main categories, from the literature were outlined of which respondents had to indicate the importance of these attributes on a five-point Likert scale. The seven categories were project characteristics; project documents; company characteristics; bidding situation; economic situation; client characteristics; and consultant characteristic. The most important attributes, identified using the Hungarian method, were chosen to establish the model. In concluding on the application of the model, the authors noted that "…although the model has many advantages, it also has some limitations in practical application: it is relatively time consuming and expensive to collect the fuzzy inference rules; and the structure of the model is relatively more complicated than other models." Based on his working experience, Connolly (2006) discussed three areas in relation to Liu and Ling’s model: the structure of markup; black box versus FNN approach to selecting markup; and the opportunity for further applied research in relation to the model. Connolly (2006) discussed three separate components of markup from his construction bidding experience: contingency on the cost of the works; contingency on the cost of risk; and price of profit. In short, the author described how contingency can be allocated as: (1) the cost of the works (calculating the work quantities (materials, labour and equipment) before using a historical database to calculate a margin of error around the accuracy of the quantities. A probability range is then used to decide a contingency on the cost of works); (2) the cost of risk (the risk attributes can alternatively be analysed using probability functions or another approach); and (3) the price of profit (allocating a contingency for the risk). In comparison to the sophisticated nature of the authors’ model, Connolly explained, from practical experience, that "…costing of risk is less mathematically rigorous and seems to be a good candidate for further investigation into use of the authors’ FNN application… However, [instead], profit truly remains the black box of bid pricing." Tah and Carr (2000) presented another model for assessing construction project risk at the tender stage using fuzzy logic. The authors described the model as "a methodology for evaluating the risk exposure, considering the consequences in terms of time, cost, quality, and safety performance measures of a project based on fuzzy estimates of the risk components." The model development was based on a risk classification scheme, i.e. hierarchical risk breakdown structure, which separated risks that ought to be priced into those related to the management of internal resources and those that are prevalent in the external environment. According to the authors, "The risk assessment process requires an assessment of the probability or likelihood of the risk and the impact." Hence, the model is based on the fundamental PI model, which is earlier on described to have major dangers in the application of its outputs. Fuzzy logic was used to develop a common language for describing and quantifying likelihood and severity of project risks. Likelihood was modelled on a seven-point scale: very very high; very high; high; medium; low; very low; and very very low. Severity was modelled on a five-point scale: very high; high; medium; low; and very low. Using the fuzzy risk assessment model, fuzzy rules are applied to arrive at a risk value for a project. However, as at the time of its introduction, it was yet to be seen precisely how the model may be used in practice. According to the authors’ concluding statements, "…discussions are currently taking place with practitioners to determine the best way to implement such a system in practice, and to develop and validate further the concepts proposed." Indeed, no empirical work informed the model development, and so it was difficult for even the authors to know whether the model was realistic. Zeng et al. (2007) introduced another fuzzy based decision making methodology to construction project risk assessment. In justifying the reason for a new fuzzy reasoning risk assessment model, the authors said: "Many risk assessment techniques currently used in the UK construction industry are comparatively mature, such as Fault Tree Analysis, Event Tree Analysis, Monte Carlo Analysis, Scenario Planning, Sensitivity Analysis, Failure Mode and Effects Analysis, Programme Evaluation and Review Technique. Nevertheless, for effective applications of these sophisticated quantitative techniques, high quality data are a prerequisite. Regrettably, such data are hard to obtain or even have not existed in the construction industry. Moreover, they are difficult to address the uncertainties and subjectivities associated with construction activities. It is therefore essential to develop new risk analysis methods to identify and assess construction risks in an acceptable way where any risk information produced is processed and reliably applied to decision making in the project management." This shows that the authors probably did not consider earlier studies like Kangari and Riggs (1989), Tah et al. (1993), Paek et al. (1993) and Tah and Carr (2000), who all alluded to the same problem in relation to the data required for construction risk analysis, to be inadequate. Although these studies are also based on the same fuzzy sets modelling technique, they are not cited. Therefore, it is hard to see a sufficient connection of the work to the existing literature. According to the authors, "Risk magnitude (RM) can usually be assessed by considering two fundamental risk parameters, risk likelihood (RL) and risk severity (RS)". However, the problem with the fundamental PI model, as pointed out by Williams (1996) and the additional effort by Ward (1999) to address the problem are not taken account of in developing the model. Since those studies were published in the same journal, it is again hard to see how this work builds on the existing literature. Zhi’s (1995) work, which led to the technical note from Williams (1996) regarding the error in using the PI model nominally is, however, cited. The algorithm of the risk assessment model based on fuzzy reasoning and AHP consisted of five phases: preliminary phase, measurement of FI phase, measurement of RL and RS phase, fuzzy inference phase and output modification phase. The model helps to go through all the stage through a complex mathematical process to obtain a risk value which should guide the project team in making the risk response decision. 3.6.4 Underlying assumptions of the analytical risk models Based on the examination of proposed analytical approaches in the literature for contractor’s risk analysis at the tender stage, the main underlying assumptions and propositions are summarized: 1. Some of the analytical approaches for example, Al-Bahar and Crandall (1990), Tah and Carr (2000) and Zeng et al. (2007) assume that the intuitive and experiential-based approach used by contractors does not form an adequate professional and objective basis for serious project management decisions. 2. The classical proposition of most analytical approaches for example, Tah et al. (1993) and Paek et al. (1993) is that based on the evaluated risk value of a project, a risk premium should be included in the bidding price to cover risks. 3. Some risk-pricing models for example, Tah et al. (1993) and Tah and Carr (2000) assume that the direct cost estimate is risk-free. These models hardly take account of the tacit and intuitive contingencies that estimators tend to include in quantities and unit rates when building up estimates. 4. Most of the risk-pricing models examined are mainly systematic, rational, or logical in nature. Hence, they assume that the pricing is a systematic process. 5. All the analytical approaches examined require contractors to perform mathematical calculations in order to arrive at risk premiums. The main risk modelling methods are probability theory, fuzzy logic, and neural networks. 6. Most of the analytical approaches examined do not clearly specify the stages of the tender process where the models should be used, and who specifically should perform the analysis, and how subjective judgement should be taken into account. They seem to assume bidding is an event rather than a process. 7. Most of the analytical approaches examined seem to assume a single risk premium for the internal risk. They do not distinguish between the different types of internal risks that contractors may include in a bid. 8. Most analytical approaches do not recognize and specify roles for bid team members in relation to how risk should be approached in the bidding process. 9. Apart from actual cost and profit margin, most analytical approaches tend not to take account of other factors such as value that also affect bidding price. 10. Most analytical approaches tend to assume that risk premium is a line item in the bid. They tend not to acknowledge that contractors may approach risk by associating it with the individual items that build up the bidding price. 11. The analytical approaches examined prescribe a three- step process for approaching risk in the bidding process: risk identification; risk analysis and evaluation (modelling of uncertainty); and contingency allocation. 12. The analytical approaches examined are based on project risk management theory which assumes that the two basic data required for assessing risk should be the probability of occurrence and potential impact. In Chapter Ten, these underlying assumptions and propositions will be compared to the empirical account of what contractors actually do about risks when pricing bids in order to identify any areas of significant difference. 3.6.5 Critiques and shortcomings of analytical risk models Several studies of contractors provide evidence and reasons why contractors rarely use the analytical risk models that have proliferated in the literature. In separate research studies involving more than 30 contractors each, Akintoye and MacLeod (1997) and Ahmed et al. (2002) identified the following problems contractors face when confronted with project risk analysis models: 1. lack of familiarity with the techniques; 2. difficulty of applying them due to the degree of sophistication; 3. doubts regarding suitability of the techniques to construction; 4. the majority of project risks relating to the contract or construction process are fairly subjective and better handled using experience from previous projects; 5. application of the techniques are time consuming; 6. the size of most construction project does not warrant their use; 7. potential benefits of the techniques are difficult to see; and 8. most techniques require sound historical data which is often difficult to gather. Smith and Bohn (1999: 102) criticized analytical risk models for their complexity of analysis, which often limits the number of variables, or project application. Also, the risks encountered in the flow of information are not modelled. The authors’ interviews with 12 US contractors showed that they consider market competition as an overriding concern when pricing work. However, most tables of risk factors and analytical models hardly address this, although it can be related to concerns that influenced the serious incorporation of micro-economic concepts into the fundamental Friedman (1956) and Gates (1967) bidding models for optimum markup estimation to make them more applicable in reality. Here, one major shortcoming with most of the 60+ models identified is that they were not derived from the kind of information that is commonly used in practice, apparently. They are essentially analytically-derived models. Hence, this was considered to be a fundamental problem. Not to mention the sophistication involved, the propositions hardly incorporate the reality that market premium may, in fact, wipe risk premium especially as estimators deal with costs whereas Directors deal with premiums. Thus, the analytical approaches are insensitive to the commercial exigencies of bidding practice. Clearly, some studies have elicited reasons for the low take-up of models. However, none has gone as far as examining what contractors actually do when they put a bid together. In short, the literature review revealed that: • There is no comprehensive literature that empirically describes the whole process of how contractors actually put a bid together from start to finish; • There is no detailed empirical account on how contractors actually take account of risks when calculating prices for their bids; and • There is no evidence that pricing is indeed systematic in nature. This combination of shortcomings in current approaches to risk analysis and pricing has led to the basic research question underpinning this study: how do contractors take account of risk when they are calculating their bids for construction work? Without empirical work which sufficiently explains what actually happens in practice, which would inform or justify the development of a new approach, it would be hard to align theory and practice. Thus, the following questions should be addressed: • How do contractors actually arrive at a bidding price? • What stages of the bidding process does risk apportionment occur? • How do contractors apportion the risk? • What factors influence the risk levels apportioned? • What kinds of information do contractors process to price risks? • How is the information processed, and how much time is spent processing it? • What activities and roles are involved in pricing risks? • To what extent is the bid-pricing process rational or systematic? • To what extent are analytical risk models used by contractors in practice? Clearly, the research required to find answers to the questions raised by this critique needs to be designed to observe what contractors do, rather than merely asking questions based on what the literature reports. The next Chapter describes how an appropriate research strategy was chosen for ascertaining what contractors actually do about risks in the whole process of putting together their bids for construction work. 3.7 Problem statement Construction projects are risky. However, in practice, contractors experience difficulties in deciding appropriate allowances for the risk when pricing work. Formal and analytical approaches that contractors can use to assess risks during tender preparation for the purpose of allocating risk contingencies have proliferated in recent years. However, they are rarely taken up in practice. Most of the past research work has focused on introducing yet more models but the low take up of this proliferation of models by contractors indicates that more models would not necessarily help. A better understanding is needed of how contractors arrive at a bidding price, and how, and in what circumstances, that price is influenced by the apportionment of risk.
US